Inspect
Read the current account, zone and deployment state.
Artificial Mind Hive // Cloudflare Ops MCP
AI can plan it.You approve it.
Walrus Tusk connects Codex, Claude, ChatGPT and other MCP-capable AI agents to Cloudflare operations—with dry runs, explicit approval, deployment checks and an audit trail between intent and production.
Open source · MIT licensed · Built for scoped Cloudflare access · Human approval stays in the loop
01Inspect first
02Dry-run default
03Approve writes
04Verify release
05Keep the trail
Safe operations
A control plane between the model and your cloud.
Give an agent useful Cloudflare tools without turning a prompt into an unchecked production change.
Plan
Turn the request into a bounded, reviewable operation.
Dry run
Show the intended diff before touching production.
Approve
Require an explicit human signal for state-changing work.
Verify
Check the deployed result and preserve the audit record.
What it handles
Cloudflare work, made legible to agents.
A practical tool layer for developers, platform teams, AI studios and operators working at the edge.
Workers & Pages
Prepare and verify edge application releases without hiding the deployment boundary.
DNS & domains
Inspect zones and plan record changes with a clear before-and-after review.
Security controls
Work with scoped access and explicit intent around sensitive configuration.
AI operations
Connect MCP-capable assistants to repeatable Cloudflare workflows.
Release proof
Verify the public result after apply instead of treating command success as proof.
Audit trail
Keep the plan, approval and outcome together for later review.
Search + agent ecosystem
Built for the AI tools developers already use.
Walrus Tusk is an independent Cloudflare operations layer—not a Cloudflare, OpenAI, Anthropic, Google or Microsoft product.
CloudflareModel Context ProtocolOpenAI CodexClaudeChatGPTGoogle GeminiEdge AICloudflare WorkersWranglerNode.jsAI agentsDeveloper automation
Regional pages
One tool. Local context.
Country pages use clean URLs and language suited to the builders and teams in each market.
Start with the code
Put a gate between “do it” and “done.”
Inspect the open-source project, review the permissions and connect Walrus Tusk to an MCP-capable agent.